feat: Report drifted resources in ClusterSummary

[gianlucam76]

Fixes #1940

Two changes that go together: say which resources drifted, and give pull mode the same per-chart behavior as push mode.

drift-detection now lists the drifted resources in ResourceSummary.status.driftedResources. addon-controller records them in the ClusterSummary:

• New FeatureSummary.lastDrift: detectedTime, the drifted resources (group, kind, namespace, name, and for Helm the release name and namespace), and a truncated flag when more than 20 drifted.
• Kept after the drift is corrected and replaced by the next drift. Only set for ContinuousWithDriftDetection.
• After a drift is recorded, driftedResources and driftedResourcesTruncated are cleared on the ResourceSummary. drift-detection only appends and skips a resource already listed, so without this the detectedTime stays pinned to the first drift.

In pull mode every chart of a profile was staged and re-applied by sveltos-applier on any change or drift. Now:

• HelmChartSummary tracks appliedContentHash (what sveltos-applier last confirmed as applied) and stagedContentHash (staged, not yet confirmed). The hash covers the rendered resources and the settings that change how they are applied.
• A chart whose hash matches the applied one, and that drift-detection did not report, is staged with ConfigurationBundle skipApply, so sveltos-applier leaves it alone. Skipped bundles are still reported, so the applier's cleanup and ResourceSummary stay correct.
• A chart is never skipped while a staged hash is pending, so a failed apply is retried.
• A chart staged because it yielded to a lower tier profile is always applied: that is how the owner tier is updated on its resources.
• Push mode is unchanged.